Probabilistic Inference Methods Research Articles

Bayesian inference of the spatial distribution of steel corrosion in reinforced concrete structures using corrosion-induced crack width

Observations of corrosion-induced crack widths offer crucial information about the corrosion states of steel reinforcements in reinforced concrete (RC) structures, enabling a cost-effective method for inferring corrosion states through inverse analysis. However, the uncertainty associated with the relationship between corrosion-induced cracking and steel weight loss necessitates a probabilistic inference method, especially when considering the spatial distributions of steel weight loss, which provides important information to estimate the load-bearing capacity loss of corroded RC structures. This paper proposes a Bayesian framework to infer the steel weight loss distribution in RC structures based on the observed corrosion-induced crack width. To reduce the dimensions of the Bayesian inference, a Karhunen-Loève transform is applied to extract the principal distribution features of the steel weight loss. The forward model of the Bayesian inference adopts a data-driven sequence-to-sequence transduction approach to predict corrosion-induced crack width from steel weight loss. This model incorporates a novel nonlinear convolution kernel for input encoding and a sparse polynomial chaos expansion for decoding, which proves more accurate and efficient than finite element simulations. The Hamiltonian Markov chain Monte Carlo (HMCMC) sampler is used to efficiently sample from the posterior distribution of the Bayesian inference. The case study of the proposed method demonstrated that Bayesian inference provides robust range estimation for the steel weight loss distribution, with its 95% confidence interval encompassing most observations. Additionally, the method efficiently inferred high-dimensional steel weight loss sequences up to 61 dimensions, taking advantage of the dimension reduction technique and the gradient-informed HMCMC sampler.

Reducing the Environmental Impact of Wireless Communication via Probabilistic Machine Learning

Machine learning methods are increasingly adopted in communications problems, particularly those arising in next generation wireless settings. Though seen as a key climate mitigation and societal adaptation enabler, communications related energy consumption is high and is expected to grow in future networks in spite of anticipated efficiency gains in 6G due to exponential communications traffic growth. To make meaningful climate mitigation impact in the communications sector, a mindset shift away from maximizing throughput at all cost and towards prioritizing energy efficiency is needed. Moreover, this must be adopted in both existing (without incurring further embodied carbon costs through equipment replacement) and future network infrastructure, given the long development time of mobile generations. To that end, we present summaries of two such problems, from both current and next generation network specifications, where probabilistic inference methods were used to great effect: using Bayesian parameter tuning we are able to safely reduce the energy consumption of existing hardware on a live communications network by 11% whilst maintaining operator specified performance envelopes; through spatiotemporal Gaussian process surrogate modeling we reduce the overhead in a next generation hybrid beamforming system by over 60%, greatly improving the networks' ability to target highly mobile users such as autonomous vehicles. The Bayesian paradigm is itself helpful in terms of energy usage, since training a Bayesian optimization model can require much less computation than, say, training a deep neural network.

Risk assessment model with probabilistic linguistic fuzzy inference methods for maritime piracy crime and applications

The surge in maritime piracy is becoming a deterrent to international seaborne trade. More attention should be attached to piracy risk analysis and prevention. The complex environments engendering piracy crime make it more difficult to conduct a quantitative judgment. Most of the existing research focuses on the risk assessment of piracy based on historical piracy activities. The dynamic evolution of piracy is not explored in these evaluation models. To solve the piracy evolution pattern analysis and risk control in the uncertain environment, we first propose a novel probabilistic linguistic Markov model to analyze and predict the risk states of piracy accidents. The basic definition of random states and the Markov process in the probabilistic linguistic environment are discussed. Next, we propose the probabilistic linguistic Bayesian network which not only depicts the causal relationships of the risk factors but also provides an efficient tool to calculate the transition probabilities in the Markov model. Then the comprehensive inference risk analysis methods are developed to conduct the risk assessment in the maritime piracy crime. The assessment results can effectively support the decision-making of risk control. A practical case study is conducted to validate the above methods. Sensitivity analyses and method comparisons also illustrate the performance of the proposed risk analysis model.

Information Theory With Kernel Methods

We consider the analysis of probability distributions through their associated covariance operators from reproducing kernel Hilbert spaces. We show that the von Neumann entropy and relative entropy of these operators are intimately related to the usual notions of Shannon entropy and relative entropy, and share many of their properties. They come together with efficient estimation algorithms from various oracles on the probability distributions. We also consider product spaces and show that for tensor product kernels, we can define notions of mutual information and joint entropies, which can then characterize independence perfectly, but only partially conditional independence. We finally show how these new notions of relative entropy lead to new upper-bounds on log partition functions, that can be used together with convex optimization within variational inference methods, providing a new family of probabilistic inference methods.

Probabilistic program inference in network-based epidemiological simulations.

Accurate epidemiological models require parameter estimates that account for mobility patterns and social network structure. We demonstrate the effectiveness of probabilistic programming for parameter inference in these models. We consider an agent-based simulation that represents mobility networks as degree-corrected stochastic block models, whose parameters we estimate from cell phone co-location data. We then use probabilistic program inference methods to approximate the distribution over disease transmission parameters conditioned on reported cases and deaths. Our experiments demonstrate that the resulting models improve the quality of fit in multiple geographies relative to baselines that do not model network topology.

Perplexity: evaluating transcript abundance estimation in the absence of ground truth

BackgroundThere has been rapid development of probabilistic models and inference methods for transcript abundance estimation from RNA-seq data. These models aim to accurately estimate transcript-level abundances, to account for different biases in the measurement process, and even to assess uncertainty in resulting estimates that can be propagated to subsequent analyses. The assumed accuracy of the estimates inferred by such methods underpin gene expression based analysis routinely carried out in the lab. Although hyperparameter selection is known to affect the distributions of inferred abundances (e.g. producing smooth versus sparse estimates), strategies for performing model selection in experimental data have been addressed informally at best.ResultsWe derive perplexity for evaluating abundance estimates on fragment sets directly. We adapt perplexity from the analogous metric used to evaluate language and topic models and extend the metric to carefully account for corner cases unique to RNA-seq. In experimental data, estimates with the best perplexity also best correlate with qPCR measurements. In simulated data, perplexity is well behaved and concordant with genome-wide measurements against ground truth and differential expression analysis. Furthermore, we demonstrate theoretically and experimentally that perplexity can be computed for arbitrary transcript abundance estimation models.ConclusionsAlongside the derivation and implementation of perplexity for transcript abundance estimation, our study is the first to make possible model selection for transcript abundance estimation on experimental data in the absence of ground truth.

Gain-loss-duplication models for copy number evolution on a phylogeny: Exact algorithms for computing the likelihood and its gradient

Gene gain-loss-duplication models are commonly based on continuous-time birth–death processes. Employed in a phylogenetic context, such models have been increasingly popular in studies of gene content evolution across multiple genomes. While the applications are becoming more varied and demanding, bioinformatics methods for probabilistic inference on copy numbers (or integer-valued evolutionary characters, in general) are scarce.We describe a flexible probabilistic framework for phylogenetic gain-loss-duplication models. The framework is based on a novel elementary representation by dependent random variables with well-characterized conditional distributions: binomial, Pólya (negative binomial), and Poisson.The corresponding graphical model yields exact numerical procedures for computing the likelihood and the posterior distribution of ancestral copy numbers. The resulting algorithms take quadratic time in the total number of copies. In addition, we show how the likelihood gradient can be computed by a linear-time algorithm.

Bayesian inference for short-time risk evaluation considering time-varying contingencies

To evaluate the short-time risk results of power systems under time-varying contingencies, a probabilistic Bayesian inference method is implemented here to construct the bridge between corrective behaviours representing rescheduling processes and steady-state scheduling conditions. The novelty lies in an efficient calculation of risk results according to given operation conditions, which is fit for online evaluation. Notably, both risk indexes, e.g., Expected-Energy-Not-Served (EENS), Lost-Of-Load-Probability (LOLP), are taken as predictors to quantify the risk of power systems, and then, these predictors can be inferred through Bayesian method. And the related case studies on the IEEE 6 bus test system are carried out to test the feasibility of the proposed model.

Recent Progress and Trend of Robot Odor Source Localization

Accurate and prompt localization of hazardous release sources ensures a timely emergency response to potential damages. Robot odor source localization is the study of locating the release source based on mobile robots in real‐time, which has been experiencing rapid growth in recent years. This paper presents a review of this field, focusing on recent progress and development trends. Research works are grouped according to method principles, leading to four categories: reactive methods, heuristic searching methods, probabilistic inference methods, and learning methods. The odor‐source direction and distance prediction via in situ sensing and developments of simulators are also listed separately for discussion. Several outlooks are highlighted at the end. Along with the developments of related fields and practical needs, we believe this field is bound to keep on flourishing in the future. © 2021 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

СОВРЕМЕННЫЕ ВЫЧИСЛИТЕЛЬНЫЕ СТРАТЕГИИ ДЛЯ ВЫВОДА БЕЛКОВ В ПРОТЕОМИКЕ ДРОБОВИКА

Today, shotgun proteomics is a powerful approach to characterize proteomes in biological samples. Unlike the top-down proteomics strategy, shotgun proteomics is characterized by high separation efficiency and mass spectral sensitivity. At the same time, it places higher demands on the computational and statistical methods required for peptide identification, protein identification, and label-free quantification. The main purpose of shotgun proteomics is to identify the shape and amount of each protein by combining liquid chromatography with tandem mass spectrometry. The analysis and interpretation of experimental data is the final and most important stage in proteomics; they also generate a large number of problems that require complex computational solutions. One of the most important tasks, of course, is the identification of proteins present in the experimental sample. As a rule, this task is divided into two main components: the stage of assigning experimental tandem mass spectra to peptides obtained from the protein database, and the stage of comparing peptides with proteins and quantitative assessment of the reliability of the identified proteins. It is also worth considering that the assessment of the reliability of the data obtained can be a separate, no less important and complex task. In this article, we propose to consider protein identification only as a problem of statistical inference, and also describe a number of methods that can be used to solve it. We classify the existing approaches into (1) rule-based methods, (2) combinatorial optimization methods, and (3) probabilistic inference methods. Integer programming and Bayesian inference frameworks are used to represent methods. We also discuss the main problems of protein identification and suggest possible solutions to these problems.

An online operating performance evaluation approach using probabilistic fuzzy theory for chemical processes with uncertainties

Operating performance evaluation (OPE) has been playing an essential role to ensure the effective operations of chemical processes. However, most of previous research focused on the deterministic evaluation strategies, without consideration of uncertainties in the evaluation indicators of OPE. Based on probabilistic fuzzy theory, an online OPE scheme is proposed by considering the uncertainties in chemical processes. In the modeling step, on the basis of just-in-time learning and probabilistic principal component regression, a prediction model is proposed and applied to estimate the probability distribution of the evaluation indicators in real time; and a weighted cosine Mahalanobis-Taguchi system for variable selection is developed to improve the prediction accuracy of the evaluation indicators. In the evaluation step, a probabilistic fuzzy inference method is proposed to improve the accuracy of evaluation results by considering the uncertainty of evaluation indicators. The effectiveness of the proposed approach is finally tested on an industrial hydrocracking process.

Robust Least-Squares Support Vector Machine Using Probabilistic Inference.

The least-square support vector machine (LS-SVM) has been deeply studied in the machine-learning field and widely applied on a great deal of occasions. A disadvantage is that it is less efficient in dealing with the non-Gaussian noise. In this article, a novel probabilistic LS-SVM is proposed to enhance the modeling reliability even data contaminated by the non-Gaussian noise. The stochastic effect of noise on the kernel function and the regularization parameter is first analyzed and estimated. On the basis of this, a new objective function is constructed under a probabilistic sense. A probabilistic inference method is then developed to construct the distribution of the model parameter, including distribution estimation of both the kernel function and the regularization parameter from data. Using this distribution information, a solving strategy is then developed for this new objective function. Different from the original LS-SVM that uses a deterministic scenario approach to gain the model, the proposed method builds the distribution relation between the model and noise and makes use of this distribution information in the process of modeling; thus, it is more robust for modeling of noise data. The effectiveness of the proposed probabilistic LS-SVM is demonstrated by using both artificial and real cases.

Complexity, Power Laws and a Humean Argument in Risk Management: The Fundamental Inadequacy of Probability Theory as a Foundation for Modeling Complex Risk in Banking

Whenever risk managers are confronted with deep uncertainty and organized complexity, probabilistic inference methods are used. These seem able to allow for crisp inputs and precise results. However, as has been noted by several thinkers (e.g., von Hayek in Studies in philosophy, politics and economics, Routledge, London, 1967; Weaver in American Scientist, 36: 536–544, 1948), such methods cannot be used effectively in such situations. This might basically sound like old wine in a new bottle and, in fact, objections to, and limitations of conventional, i.e., probabilistic risk modeling are anything but unheard of in the literature of banking and finance. However, this paper introduces for the first time an argument, inspired by reflections on the old riddle of induction, from which those shortcomings of the limited suitability of probability can be derived. It demonstrates that any choice of a particular probability distribution for a given risk management purpose is necessarily arbitrary, i.e., it is not grounded in the data but in the choices of the statistician or risk manager, and cannot be justified by appealing to something more objective. Thereby, this paper unmasks the illusion that financial data and extreme losses are well-described by non-standard probability functions such as power laws that have been embraced at the expense of bell curves in the aftermath of the global financial crisis of 2008. Moreover, although we do not propose a positive solution, we believe that articulating the real, and as yet unnoticed, source of the problem is a key step towards developing a principle and tractable response.

Properly initialized Bayesian Network for decision making leveraging random forest

Bayesian network is one of major methods for probabilistic inference among items. But if it contains particular targeting node and other explanatory nodes for decision making, for example how to select suitable appealing keywords to make customers like a product, edges around the target should be counted with more importance than those among others while constructing the network. In order to achieve this adjustment, this study proposes to configure initial state consisting of a few nodes and their edges connected with the target. The initial state is obtained by leveraging Random forest which is a proven method for decision making. Initial nodes are extracted by measuring mean decrease of Gini coefficient calculated with decision trees of Random forest. Initial edges are designated by comparing Kullback-Leibler divergences of conditional probability distribution among nodes which are corresponding to edge directions. Through an actual experiment, this method is confirmed to be effective for adjusting Bayesian network in decision making. This approach is especially useful for business scenes, such as selecting preferable keywords for appealing products over SNS.

Home service robot task planning using semantic knowledge and probabilistic inference

In the face of unstructured home environment, home service robots are inevitably confronted with uncertainty and incompleteness of environment information. How to make the home service robot obtain enough environment information and plan a discrete sequence of actions through task planning is the key problem of robot intelligence. In this paper, a hierarchical task network based on semantic knowledge and probabilistic inference method is proposed. We use the object location ontology, the location relation between dynamic and static objects to build semantic knowledge of home environment, and build the probability model between dynamic and static objects, as well as between static objects and home scenes. The location of the object is determined by the semantic knowledge and the probability model. Hierarchical task network is selected as an engine of task planner, which can be provided with the location information to improve the autonomy and effectiveness of robot task planning. In order to prevent task execution failure and enhance the adaptability of robot to unstructured home environment, a mechanism of task execution diagnosis and replanning is designed. Experimental results in simulation and real home environment demonstrate that our method can effectively improve the performance of service robot task planning and generate better task execution sequence.

Risk assessment of supply-chain systems: a probabilistic inference method

ABSTRACT The probability of occurrence, occurrence time and holding time make the estimation of the multiresource, multidimensional, transmissible and dynamic characteristics of supply-chain risk more valuable to achieving the goal of risk assessment. A dynamic Bayesian network model of supply-chain risk is constructed to describe the property of supply-chain risk; the Bayesian inference tool is then used to estimate the corresponding parameters by maximum likelihood and inference for supply-chain risks. It is concluded that supply-chain risk changes with time and would converge at a certain stable interval, occurrence time and holding time satisfy several Poisson processes.

Ptype: probabilistic type inference

Type inference refers to the task of inferring the data type of a given column of data. Current approaches often fail when data contains missing data and anomalies, which are found commonly in real-world data sets. In this paper, we propose ptype, a probabilistic robust type inference method that allows us to detect such entries, and infer data types. We further show that the proposed method outperforms existing methods.

Effective elastic thickness over the Chinese mainland and surroundings estimated from a joint inversion of Bouguer admittance and coherence

The effective elastic thickness (Te) of the lithosphere reflects the long-term flexural strength and provides mechanical information about the evolution and deformation of the lithosphere. The Chinese mainland and its surroundings (CMS) comprise sharply contrasting tectonic domains and involve a complex evolution history. In this paper, we apply a joint inversion of real admittance and coherence calculated using the continuous planar wavelet transform to estimate the spatial variations of Te in the CMS based on topography and Bouguer gravity anomaly data. A probabilistic Bayesian inference method is used to show the statistical distribution of the model parameters in specific cases, demonstrating the accuracy and stability of the joint inversion. Old stable cratons and basins (e.g., Indian Plate, Tarim Basin, Junggar Basin and Lower Yangtze) mostly correspond to high Te (>50 km), whereas reactivated Mongolian Plateau, eastern North China Craton, Cathaysian Block, continental margins and young oceanic lithospheres tend to display low Te (<25 km). Moreover, Te correlates remarkably well with lithospheric thermal structure; areas with high Te are mostly characterized by low heat flow and large Curie-point depth, and vice versa. We conclude that the geothermal state plays an important role in controlling the continental lithospheric strength, although in young tectonic domains, such as Tibetan Plateau and the East China Sea Basin, lithospheric structure and tectonics have strong effects on lithospheric strength.

Bidirectional General Graphs for inference. Principles and implications for medicine

Probabilistic inference methods require a more general and realistic description of the world as a Bidirectional General Graph (BGG). While in its original form the Bayes Net (BN) has been promoted as a predictive tool, it is more immediately a way of testing a hypothesis or model about interactions in a system usually considered on a causal basis. Once established, the model can be used in a predictive way, but the problem here is that for a traditional BN the hypotheses or models that can be formed are limited to the Directed Acyclic Graph (DAG) by definition. Three interrelated features are highlighted that represent deficiencies of the DAG which are corrected by conversion to a method based on a BGG: (i) lack of intrinsic representation of coherence by Bayes' rule, (ii) relatedly the need to consider interdependence in parent nodes, and (iii) the need for management of a property called recurrence. These deficiencies can represent large errors in absolute estimates of probabilities, and while relative and renormalized probabilities ameliorate that, they can often make much of a net superfluous through cancelations by division. The Hyperbolic Dirac Net (HDN) based on Dirac's quantum mechanics is a solution that led naturally to avoiding these deficiencies. It encodes bidirectional probabilities in an h-complex value rediscovered by Dirac, i.e. with the imaginary number h such that hh=+1. Properties of the HDN described previously are reviewed (though emphasis is on descriptions in familiar probability terms), the issue of recurrence is introduced, methods of construction are simplified, and the severity of the quantitative differences between BNs and analogous HDNs are exemplified. There is also discussion of how results compare with other approaches in practice.

The Hybrid parameter learning algorithm development for dynamic Bayesian network in the context of the Metropolis — Hastings approach

Bayesian networks are an efficient tool for analyzing complex stochastic processes. The generic methods of analysis based on Bayesian networks are probabilistic inference procedures that aimed at solving filtering, prediction and smoothing problems. An essential stage in the probabilistic inference methods implementation is network structure and parameters learning carried out on the special training samples obtained as a result of passive observation of the analyzed process and/or active experiment. Effectiveness of the Bayesian networks parameters learning algorithms directly depends on the preprocessing and organization of the training sample processing, which optimizes the classical learning algorithms capabilities. In the context of research introduced in this article, a hybrid algorithm for the parameters learning of dynamic Bayesian networks with observed and hidden variables has been developed. Algorithm allows increasing the efficiency of probabilities calculation procedure for the initial state and the probabilities of transition between the networks slices involving the variety of special methods for organizing training sample preprocessing. The questions of convergence of the proposed algorithm are also considered.